Hyperagent is the system of agents that does real work, learns how your organization operates, and deploys across your entire team. By the team at Airtable.
We wanted better design fundamentals from our agents.
So we fed them this 162-page pdf on designing with a grid system.
Now our agents use code to adhere to a grid and design beautiful layouts.
Example + skill below 👇
New skill on marketplace: Roast My Idea by @nateherk
Convene a council of five independent personas who debate your idea from every angle.
Built to overcome the sycophantic tendencies of some models.
Get the skill below ⬇️
New in @hyperagentapp:
• Model-agnostic. Claude, GPT, Gemini, GLM, Kimi.
• Orchestrators can hand off to specialist agents.
• Marketplace is live for agents and skills.
• MCP server, so your agents plug in anywhere.
• $500 in credits if you're switching from OpenClaw
GPT-5.6 is a viable alternative to Fable for most knowledge work.
GPT-5.6's writing is consistently well-structured and easy to read. Fable's prose can be exhausting.
5.6 also makes fresh design choices with clean hierarchy. Fable can struggle to maintain visual relationships
You have a new job.
Your role is now to oversee your fleet of agents as they work autonomously, set their guardrails, cultivate their shared learning layer, and intervene only when needed.
Here is the form factor we think that takes, shown with a real example:
Fable 5 and Sonnet 5 really need each other
We ran every Claude model on Hyperagent to build a creative data viz about goals scored at the World Cup
Here's how they did 👇
Agent browser use is still so underrated.
If your agents can't actually see webpages, they're walking around blindfolded. So much high-value work cannot be done through APIs alone.
Hyperagent runs @browserbase to visually navigate the web live.
Mutliplayer agents are the new default.
Singleplayer is fine for personal productivity, but you'll only get org-level impact if your team owns the agents together.
We just added Team-owned Agents to @hyperagentapp to solve this
Fashion brands are turning basic studio shots into editorial lookbooks for <$30
Hyperagent acts as art director, then calls Nano Banana and GPT-Image-2 to generate images.
Product details are respected from the real photography, down to the exact patterns and textures on a
An open-weight model just beat the frontier flagships at competitive intelligence - for a fraction of the price.
We added four open models to Hyperagent - GLM 5.2, DeepSeek V4, Kimi K2.6, and Qwen 3.7 - and tested them on the work our users actually run: prospecting, competitive
DeepSeek V4 is built for heavy reasoning, and that's where it delivered.
Pointed at a large public research dataset, it surfaced a trend most would miss: the U.S. has lost its long lead in scientific output - and that the quality gap between countries closed years before the
Qwen 3.7 was the value standout.
The data stories it produced - clean charts, real imagery, a clear narrative - cost 13 to 20 cents per run, well below anything else producing work at that level. ($0.50)
See it here: hyperagent.com/s/DIIUcCQCFXCj…
GLM 5.2 went up against the frontier flagships on competitive intelligence and beat them, at roughly a sixth of the cost.
It was also the only model that questioned a competitor's marketing claim instead of repeating it - taking an advertised "14x conversions" and showing the